Sensorless Control of Permanent Magnet Machine for NASA

NASA / TM--2002-211726 IEC E C-2002-20072

Sensorless Control of Permanent Magnet

Machine for NASA Flywheel

Technology Development

Barbara H. Kearny

Glenn Research Ce:nter, Cleveland, Ohio

Peter E. Kascak

Ohio Aerospace Insti.tute, Brook Park, Ohio

July 2002

The NASA STI Program Office... in Profile

Since its founding, NASA has been dedicated to

the advancement of aeronautics and spacescience. Tile NASA Scientific and Technical

Information (STI) Program Office plays a key part

in helping NASA maintain this important role.

Tile NASA STI Program Office is operated by

Langley Research Center, the Lead Center forNASA's scientific and technical information. The

NASA STI Program Office provides access to the

NASA STI Database, the largest collection of

aeronautical and space science STI in file world.

The Program Office is also NASA's institutional

medlanism for disseminating the results of its

researd3 and development acti vities. These results

are published by NASA in the NASA STI Report

Series, which includes the following report types:

TECHNICAL PUBLICATION. Reports of

completed :research or a major significant

phase of research that present the results of

NASA programs and include extensive data

or theoretical analysis. Includes compilations

of significant scientific and technical data and

information deemed to be of continuing

reference value. NASA's counterpart of peer-

reviewed formal professional papers but

has less stringent limitations on manuscript

lengfl3 and extent of graphic presentations.

TECHNICAL MEMORANDUM. Scientific

and tedmical findings that are pre, liminary or

of specialized interest, e.g., quick release

reports, working papers, and bibliographiesthat contain minimal annotation. Does not

contain extensive analysis.

CONTRACTOR REPORT. Scientific and

technical findings by NASA-sponsored

contractors and grantees.

CONFERENCE PUBLICATION. Collected

papers from scientific and technical

conferences, symposia, seminars, or other

meetings sponsored or cosponsored byNASA.

SPECIAL PUBLICATION. Scientific,

technical, or historical information from

NASA programs, projects, and missions,

often concerned with subjects having

substantial public interest.

TECHNICAL TRANSLATION. English-

language translations of foreign scientific

and technical material pertinent to NASA'smission.

Specialized services that complement the STI

Program Office's diverse offerings include

creating custom thesauri, building customized

data bases, organizing and publishing research

results.., even providing videos.

For more information about the NASA STI

Program Office, see the following:

® Access the NASASTI Program Home Page

at http:lhuww.sti.nasa.gov

® E-mail your question via the Intemet to

[email protected]

* Fax your question to the NASA Access

Help Desk at 301-621-0134

* Telephone the NASA Access Help Desk at301-621-0390

Write to:

NASA Access Help Desk

NASA Center for AeroSpace Information7121 Standard Drive

Hanover, MD 21076

NASA / TM--2002-21_ 1.726 IEC E C-2002-20072

Sensorless Control of Permanent Magnet

Machine for NASA Flywheel


Barbara H. Kearny

Glenn. Research Ce:nter, Clevel.and, Ohio

Peter E. Kascak

Ohio Aerospace Institute, Brook Park, Olnio

Prepared for the

37th Intersociety En.ergy Conversion Engineering Conference

sponsored by the Institute of Electrical and Electronics Engineers,

Electron Devices Society

Washington., DC, July 28-August 2, 2002

National Aeronautics and

Spa ce Ad.minis tration

Gleam Research Center

July 2002

Acknowledgments

The authors would like to acknowledge the helpful comments of Dr. M. David Kankam during his review of this

paper and the valuable assistance of Mr. Tim Dever in collecting the data for this paper.

NASA Center for Aerospace Information71121Standard Drive

Hanover, MD 211076

Available frorn

National Technical Information Service

5285 Port Royal RoadSpringfield, VA 22100

Available electronical] y at http://gltrs.grc.nasa.gov/GLTRS

IECEC 2002 20072

SENSORLESS CONTROL OF PERMANENT MAGNET MACHINE FOR NASA FLYWHEEL

TECHNOLOGY DEVELOPMENT

Barbara H. Kenny

National Aeronautics and Space AdministrationGlenn Research CenterCleveland, Ohio 44135

216---433---6289

barbara.kenny @grc.nasa.gov

Peter E. Kascak

Ohio Aerospace InstituteBrook Park, Ohio 4.4.142

[email protected]

ABSTRACT

This paper describes the position sensorlessalgorithms presently used in the motor control for theNASA "in-house" development work of the flywheel energystorage system. At zero and low speeds a signalinjection technique, the self-sensing method, is used todetermine rotor position. At higher speeds, an open loopestimate of the back EMF of the machine is made todetermine the rotor position. At start up, the rotor is setto a known position by commanding dc into one of thephase windings. Experimental results up to 52,000 rpmare presented.

INTRODUCTION

NASA is in the process of developing the technologynecessary to make flywheels attractive for use as energystorage devices in space applications. One of thenecessary components is the motor/generator and itsassociated control (Kenny, 2001). The permanentmagnet (PM) machine is an attractive choice for themotor/generator because of its high power density andefficiency. The PM machine can be accurately controlledwith a fast response and low torque ripple through theuse of the vector control algorithm (field orientation)(Krishnan,1999). However, field orientation requires rotorposition information with a resolution on the order of a fewdegrees (Jahns, 1994). Position feedback with thisresolution for PM machines is typically provided by usingeither an encoder or a resolver. However there are speedlimitations on these devices and even in speed rangeswhere position feedback devices are feasible, asensorless approach still has the advantage of reducingthe complexity of the system by eliminating a hardwarecomponent and its associated failure modes, volume andmass.

A variety of sensorless techniques have beenproposed in the literature and it is still an active field ofresearch (Johnson, 1999). This work combines two ofthe techniques to achieve sensorless control at allspeeds: the self-sensing technique for low and zerospeed operation (Jansen, 1995) and a back EMF basedmethod for higher speeds. A similar combination isdescribed in (Patel, 2001).

NOMENCLATURE

Lq is the q-axis machine inductance, henries.

L d is the d-axis machine inductance, henries.

x represents a q-axis scalar quantity where f can befqy

voltage, v, current, i, or flux, ;L; x can be the rotorreference frame, r, or the stator reference frame, s, orthe negative sequence reference frame, neg; and y iseither the stator variable, s, or the rotor variable, r.

fc_y represents a d-axis scalar quantity where f can be


fqdyX represents the d-q vector quantity where f can be


A superscript asterisk, *, represents a commandedvalue.

A carrot above a symbol, ^, represents an estimatedquantity.

NASA/T_2002-211726 1

s •

Vqds c is the commanded high frequency carrier signalvoltage vector for the self-sensing technique.

p is the derivative operator, d/dt.

Or is the angle between the stator q-axis and the rotor

q- axis, radians.

;gaf is the flux linkage due to the rotor magnets, volt-sec

% is the electrical rotor speed, radians/second.

coc is the carrier signal frequency, radians/second.

c% is the fundamental electrical excitation frequency,radians/second.

SELF-SENSING TECHNIQUE

The self-sensing method of position estimation in ACmachines was first described in (Jansen, 1995). It is asensorless technique that relies on a magnetic saliencypresent in the machine to provide position-dependentinformation which can be extracted with the appropriatesignal processing. Unlike the back EMF techniques, themethod does not rely on fundamental excitation of themachine so position estimation is possible at low andzero speeds. However, the method does require anadditional signal into the machine which will lead toadditional losses.

A necessary requirement of the self-sensingtechnique is a known magnetic saliency in the machineso that a position (or flux) dependent inductance results.Most permanent magnet machines inherently have somemagnetic saliency with interior permanent magnet

machines exhibiting the most saliency (Lq = 2 to 3 times

Ld) and surface mount permanent magnet machines

exhibiting a very small saliency (Lq=l.O5 to 1.15 times Ld)(Krishnan,1999). The machine used in this work is a 4

pole surface mount permanent magnet machine with Lq =

139 pH and La = 116 pH. The larger the saliency, the

easier it is to track the position-dependent signal.However, even with the small saliency in this machine itis possible to successfully estimate the rotor position.

The other necessary requirement of the self-sensingtechnique is a high frequency carrier signal in addition tothe fundamental excitation. The carrier excitation signal

(or

is independent of the fundamental operation of themachine and is thus present at all speeds. The carrier

S_

signal, Vqa s c, is added in a feedforward fashion to thecommanded fundamental voltage and both aresynthesized by the PWM action of the inverter switches.The block diagram of the overall system is shown in Fig.1. The vector quantities shown in Fig. 1 are defined by(1) where f can represent current, flux or voltage. Bothscalar and vector forms are used in this paper.

fqds = fqs - Jfds (1)

A high frequency current with a position-dependentphase variation results from the injection of the carriersignal voltage and the position-dependent inductancevariation in the machine. The expression for this highfrequency current will now be derived.

The machine inductance vector, expressed in therotor reference frame, is given in (2).

, [,q0]Lqd 0 LdJ

(2)

The expression for the stator flux linkage at the carrierfrequency is shown in (3).

),as]=[LqOl[iqs1;L_sj 0 LdJ L i_sJ

(3)

The transformation between the rotor reference frame and

the stator reference frame is shown in (4)(Krish nan, 1999).

[ fqs] = [c°S(er)sin(er) ] [fqs]fc_sj L- sin(Or) cos(er)J Lfc%J (4)

Substituting (4) into (3) transforms the equations from therotor reference frame into stator frame as follows:

s*

.r*

)-'{ *SynchronouslVodsr*.r* ^ IFrameCurren,ias=-0-_I Regulator ] --I F'rame _ VDe°t°r [_] I_ L

finds _r_

/_ I --q Self_sensing i s II_I--l,_a"

'4ql_lrRo_r I ',_=s F -I filtering and I_ind_labc to dql .,,.s / positionand r'----Itransforml_gb--Ispeed estimation/I I

{ Back EMF /"

(_r _ position and)eed estimation

L ..... Self-sensingor back EMF

nchro

Motc

Figure 1: SYSTEM BLOCKDIAGRAM

NASA/T_2002-211726 2

-sin(Or)cos(Or)JLX_sJLoLdJL-sin(Or)cos(Or)JLi_sJ(5)

;_qsl_-cos(Or)-sin(er)lrLq 0 lr cos(Or)sin(er)lriqs])vc%JL sin(0r) cos(0r)JL 0 LdJL-sin(0r) cos(0r)JLic_s j (6)

Thus the stationary frame expression for the inductanceis given in (7) and shows the position dependentvariation.

s [ c°s(0r) -sin(er)]rLq0 ][co_(0r_ sin(0r)1=Lqd sin(0r) cos(0r)JL 0 LdJL-sin(0r) cos(0r) j

.Y-Lqd + ALqdCOS(20F) -ALqdSin(20r) ]-ALqdSin(20r) Y-,Lqd - ALqdCOS (20r)

Lq + L d Lq - L dwhere Y_,Lqd 2 ' ALqd 2 (7)

The machine model in the stator reference frame at the

carrier frequency is given in (8).

S* .S S .SVqds c = Rslqds c ÷ Lqd plqds c (8)

Substituting (7) into (8) and neglecting the statorresistance (the stator resistance is small relative to theinductive impedance at the carrier frequency) gives therelationship between the high frequency injected carriervoltage and the resulting high frequency current.

Is" 1Vqs c

s*LVds cJ

XLqd + ALqdCOS (2Or) -ALqdSin(2Or)-ALqdSin(2Or) Y--Lqd - ALqdCOS (2Or)

l[.slplqs c

.s

JLP'd,_cJ(9)

The applied high frequency carrier signal is a balancedpolyphase voltage given by (10) where V m is the peak

value of the applied signal.

[vq:cl [Vmc°S ct 1Vc% cJ L -vmsin(c°ct) J

(10)

Substituting (10) into (9) and solving for the highfrequency current results in the following.

[iqsc] [ Icp sin(c°ct) +lcnsin(-(oct+20r) ]= (11)

Li_s cJ IcpCOS((Oct) + Icnc°S( -(Oct + 2Or)

Vm Y--Lqdwhere Icp and

C0c (y_.Lqd) 2 - ( ALqd ) 2

Vm kLqdIcn

C0c (y_.Lqd) 2 - ( ALqd ) 2

Equation (11) can be expressed in vector form by usingthe definition given in (1) as shown in (12).

Iqds c= -Jlcp ejc°ct - jlcnej(20r -cod) (12)

It can be seen from (11) and (12) that there are twocomponents to the high frequency current. Both

components are at the carrier signal frequency, coc. The

Icp term is the "positive sequence" component, that is, itis rotating in the same direction as the applied voltage

(+COc). The Icn term is the "negative sequence"

component, that is, it is rotating in the opposite direction

of the applied voltage (<%). The position information, Or,

is seen to be contained in the negative sequencecomponent.

Equation (12) represents the high frequency currentin the machine due to the application of the carrier signalvoltage and the presence of the magnetic saliency. Thetotal stator current will also contain a fundamental

component with an excitation frequency c% as shown in

(13).

iqds = leeJC%t- jlcpeJCOct- jlcnej(2Or - COct) (13)

The filtering of this current will be described next.

The positive sequence component of the current,

-jIcpe jc°ct, contains no position information nor is it

necessary for the fundamental operation of the machine.A synchronous frame filter is used to effectively removethis component as shown in Fig. 2 ("HPF" is a high passfilter).

le e +Jc°et_ jlcpe+J%t - jlcne+J(2er-%t)l_

le e +j(c%t-c%t) _ jlcp _ jlcne+j(2er_2%t)

le e +j(%t-c%t) _ jlcne+J(2er-2C%t) _L" . . _lee+J% t

- llcne+_l_20r met)

Figure 2: POSITIVE SEQUENCE SYNCHRONOUS FRAME FILTER

The remaining current is seen to consist of thefundamental component and the negative sequence

component: leejc%t- jlcnej(20r - COct) The fundamental

component is necessary to regulate the fundamentaloperation of the machine. The negative sequencecomponent is necessary to estimate the rotor position.Thus two additional filtering steps are needed as shown inFigure 3. To extract the fundamental current, a

NASA/T_2002-211726 3

/l ee+j%t - j l cn e+j(2er-%t L Fj

.-i e

Ie - jlcne+J(2er-C%t-c% t) _ ,

EFq-Jlcn e+i(2er-c%t-%t) I,

ep'eq- Jlcne+i(2er-C%t)

lee+J(c%t+c% t) _ jlcn

Ele e+j(°)et+°)ct)_

eF

e_- iIone+j(2er) y

,Elr

"_1, +jO)et

,ct Ie.IJe-_.--7.-_

J = iqdsl

Figure 3: NEGATIVE SEQUENCE SYNCHRONOUS FRAME FILTER

AND FUNDAMENTAL SYNCHRONOUS FRAME FILTER

negative sequence synchronous frame filter is used. Toextract the negative sequence current, a fundamental

synchronous frame filter is used.The position information is extracted from the

negative sequence current by using a heterodyningprocess to create an error signal which is then used in anobserver structure as shown in Figure 4. In a flywheelsystem, where the majority of the electromechanical

torque, Tern, is used to accelerate or decelerate the

flywheel, the use of the estimated torque in the observerimproves the position and speed estimate by feedingforward the acceleration information to eliminate the

position and speed estimate lag. The observer structureshown in Figure 4 can also be implemented without thetorque feedforward term which results in essentially aphase locked loop structure. Note that a speed estimateis also available in this structure and is used as feedback

in the outer speed control loop of the motor.

+z:+,

r_

Figure 4: BLOCK DIAGRAM OF SELF-SENSING POSITION AND

SPEED OBSERVER

The scalar representation of the cross product shown in^

Fig. 4 is given in (14) where (e r - er) is assumed to besmall.

l(_jlcnej(2Or) X- jlcnej(2e_) ) =Icn

.neg .neg

-i'-c e'as os(2Ar)Ids a a a- _-a'-sin(28r) = sin 2(0 r- Or) = 2(0 r - Or)Icn Icn

(14)

START-UP PROCEDURE

The rotor position angle, Or, is defined as the angle

between the stationary frame q-axis and the rotor frame

q-axis. The self-sensing method determines er based on

the inductance difference between the d- and q-axes. Themethod does not inherently differentiate between thepositive q-axis and the negative q-axis. This can beseen from (14) where the phase of the filtered current isproportional to twice the rotor position. This means that arotor position angle of 0 electrical degrees will result inthe same negative sequence current as a rotor positionangle of 180 °. If the rotor position is at 0 electricaldegrees and the self-sensing estimate locks onto 180 °,the result will be that the machine will spin in the negativedirection for a positive speed command. So an initiationtest must be done at start up to determine if the self-sensing estimate has locked onto the correct angle.

The initiation test is done as follows. First the rotor

is moved to a known position. This is done bycommanding a small dc voltage onto the stator windingswhich effectively sets up a dc magnetic field in the airgap. The rotor magnets will align with this field and thusthe rotor will move to a known position (Matsui, 1996).By commanding a small additional duty cycle to thepositive rail switch on phase A of the machine, a dc field

is established in the direction of the stator +q-axis. Therotor magnets align with this field which results in a 90 °angle between the stator q-axis and the rotor q-axis (ie.

er = 90°). The self-sensing algorithm will estimate either

er 90 ° or er -90 ° for the rotor in this position. If the

estimate is equal to -90 °, then 180 ° is added to the self-sensing estimate and that result is used in the control

algorithm for e r . If the estimate is equal to 90 ° , then

there is no offset angle added to the estimate beforebeing used in the control algorithm.

BACK EMF TECHNIQUEThe back EMF technique is based on the fact that

the spinning rotor magnets will generate a voltage (theback EMF) at the terminals of the machine. This voltagecan be integrated to find the stator flux (Wu, 1991). Fromthe stator flux vector and an estimate of the torqueangle, the rotor flux position can be estimated (Patel,2001). This will be described next.

The stator reference frame equation describing therelationship between the stator voltage, current and fluxlinkages for the fundamental excitation of the machine isgiven in (15).

S .S S

Vqd s = IqdsR s + p)Vqd s (15)

The stator flux can be estimated by integrating the statorvoltage less the IR drop as shown in (16).

)Vqds= Vqd s - IqdsR s t (16)

In the simplest implementation (used in this work),s s*

the stator voltage, Vqds, is assumed to be equal to Vqds,the commanded voltage, which is known in the controller.This is equivalent to assuming a constant dc bus voltageand no voltage loss in the inverter. Improvements in the

NASA/T_2002-211726 4

flux estimate by using either better estimates of statorvoltage or by actually measuring stator voltage arepossible and described in the literature.

Additionally, the open loop integration described by(16) was actually implemented as a low pass filter with a50 Hz bandwidth. This was found experimentally to resultin a reliable stator flux estimate at speeds above about500 rpm.

However, it is knowledge of the position of the rotor

flux, Xaf, which is necessary for proper field orientation.

This can be found by expressing the stator flux in therotor reference frame wherein the d-axis component of

the stator flux vector is aligned with the rotor flux, Xaf.

r .r

;gqs = Lqlqs (17)r .r

;Lds = Ldlds + )Laf (18)

The angle between the stator flux vector, ;gqds, and ;gaf is

the torque angle which is found from (17) and (18).

8 = tan -] Iqs. (19)

k. Ldids + ;gafJ

These relationships can be best shown graphically asin Fig. 5. For field orientation control, the d-axis of therotor reference frame is defined to be coincident with therotor flux. To transform from the stator frame to the rotor

reference frame, the rotor angle, Or, must be known. This

is the angle between the stator frame d-axis and the rotorframe d-axis (or, equivalently, the angle between thestator frame q-axis and the rotor frame q-axis) as shownin Fig. 5. Mathematically, this angle can be found from(20) as shown.

Or = tan -1 - 8 (20)

A

The estimated rotor angle, Or, is used for the

reference frame transformation as shown in Fig. 1 and is

',, _q

'", S,,,/'i

J0r

,r

II_::s1# d r

"LdiSs

qs

Vds

Figure 5: VECTOR DIAGRAM SHOWINGFLUX, CURRENT, ANGLESAND REFERENCEFRAME AXES

A

Figure 6: SPEED OBSERVER

also used as an input into a speed observer which issimilar in structure to Fig. 4 and shown in Fig. 6. Again, atorque estimate can be used to enable zero lag propertiesof the speed estimate (Lorenz, 1991).

EXPERIMENTAL RESULTS

The sensorless algorithms were implemented tocontrol this motor precisely because position feedbackwas not available. Thus it was not possible toindependently verify the exact accuracy of the positionestimates. However, a general indication of the fidelity ofthe estimates is presented here by observing the motoroperation and the position estimates under certainconditions, including successful operation to 52,000 rpm.

One method to gauge accuracy is to observe theposition estimates while the machine is freelydecelerating ("spinning down"). Under this condition, thefundamental stator current is commanded to zero and

(15) reduces to:

s .s [v;1 [p ;1 [C°r;LafCOS(0r) 1

Vqds = p/Nds = [V_J = [p;Lc_sJ = L_COr;Lafsin (er) J (21)

Thus there is a way to relate the position estimate tothe fundamental phase voltage applied to the machine.The actual phase voltages were not measured. However

S*

the commanded voltage, Vqs, is available in the controllerand was assumed to be approximately equal to the actualphase voltage, Van.

Figure 7 shows the rotor position estimates during aspin down test from 1500 rpm. It can be seen that thereis a phase difference between the two estimates. A

400

350 ,_ .ii i ;"i

_300 ' "

"_.250g

200 )":&

Oolo

II :"i iii:: /I .:

/ I" i t I"

/ I,"

/ ]I i / J

•_ i /

/ i / .."t : i i "

/ i /t ,." / ...

i • i / t •

/ ..: i: t_ ..: / .:

!/ . _ .il : ,

;' "' ;'i'"\ ;':''"/:... ,: .' ,i .?' iv:..

i :..

, self-s_nsing ,

II ": i I ..:

l ": i,. i

/ " ! I," :

/ : i ii ..'1 i:1

/

/ i ,,:

i . : ....... i i!i .. : / i/ : i

4i,' i/! • : 'I

ii , I i "

]¢! i i .I: I i:.

i . ii :..v / •back EMF

0.02 0.04 0.06 0.08 0.1time, seconds

Figure 7: er ESTIMATESFOR SPIN DOWN FROM 1500 RPM.

NASA/T_2002-211726 5

0 0.02 0.04 0.06 0.08 0.1time, seconds

Figure 8: EXPECTED AND ACTUAL Q-AXIS PHASE VOLTAGESDURINGSPIN DOWNFROM 1500 RPM

comparison of the phase shift between the actual rotorposition and the estimates can be found by using (21)and the estimated rotor angles to calculate the predicted

S S*

Vqs. The predicted voltages and Vq s are plotted in Fig. 8.It can be seen that the back EMF method estimate leads

the actual by about 44 ° while the self-sensing lags it byabout 24 ° .

The phase lead in the back EMF method is due to theimplementation of (16). A low pass filter with a bandwidthof about 50 Hz is used instead of the pure integrationdescribed in (16). The dc offsets in the measurementscause erroneous results if a pure integration is used. Apure integrator has a phase shift of 90 ° however the lowpass filter approaches a 90 ° phase shift only atfrequencies above the bandwidth. Thus the back EMFestimate improves with higher speeds. This can be seenin Figure 9 where the data from a spin down from 7500rpm is shown and the phase lead is reduced toapproximately 10 °.

The self-sensing error can be attributed to two mainfactors. First, the resistance of the machine wasneglected in the derivation of (11) and (12) and this willlead to a slight phase lag in the estimate. Second, thefiltering of the current to determine the position estimate

25

2o

15

lO

:o 5

-:Itf-1

-20I _V s-25 t q backEMF

0 0.005 0.01 0.015 0.02time, seconds

Figure 9: EXPECTED AND ACTUAL Q-AXIS PHASE VOLTAGESDURINGSPIN DOWNFROM7500 RPM

using s s esiimate using bemf estimate using s s e_timate using bemf estimate

2

nee e ina eeslm e

0 01 0 0 01 0 o2 time secondstime seconds

Figure 10: PHASE CURRENT Figure 11 : PHASE CURRENTDURINGTRANSITIONWITH NO DURINGTRANSITIONWITH

A A

OFFSET IN Or OFFSET IN Or

introduces phase errors. The relationship between theseeffects and the position estimate error is deterministicand the estimate could be improved with additional tuningand testing. However, in this application exact accuracywas not important because the self-sensing method wasused only at low speeds.

The position and speed feedback are changed fromthe self-sensing estimate to the back EMF estimatemanually, typically at 1500 rpm. Fig. 10 shows the phasecurrent during the transition with no adjustment for thephase errors of the estimates. Fig. 11 shows thetransition with both estimates adjusted appropriately. Itcan be seen that there is a spike in the current withoutthe adjustments but the effect is small and it is notnecessary to adjust for the phase errors to successfullytransition between the methods.

The speed estimate for both techniques is shown inFig. 12 at a constant speed of 1500 rpm. Both speedestimates are within +/- 1% of the commanded value.The accuracy of the speed estimate can be best seen byexamining a frequency spectrum of the current. In asynchronous 4 pole machine, the fundamental frequencyoccurs at twice the mechanical frequency. Thus at 1500rpm, the expected fundamental current frequency is 50Hz. Fig. 13 shows this to be the case. The secondharmonic oscillations in the back EMF speed estimatedecrease significantly at higher speeds, probably due tothe better position estimate. The bandwidth of the speedloop is relatively low so the speed controller does notrespond to the oscillations in the speed estimate.

1550

E 1525..................................................................................................

{, o0 .... 12

1475 .............................I self-sensing i \

back emf

145_{15 ! 0.5

time, seconds

Figure 12: ESTIMATEDROTORSPEED FOR A COMMANDEDVALUE OF 1500 RPM

NASA/T_2002-211726 6

- -_03

_025 i i

_02 i i

0_5_0_ i i

005

%"- Y0 - 20 6'_-- ¢0__A_00Atequency Hz

Figure 13: PHASE CURRENTFREQUENCY SPECTRUM, LOW

FREQUENCYRANGE

35 " " '

2

_25 t_2_15

05ZZZZZZIIII[ZZZil1 _50 1275 1300 1325 1350

tequency Hz

Figure 14: PHASE CURRENTFREQUENCY SPECTRUM, HIGH

FREQUENCY RANGE

The amount of current necessary for the self-sensingtechnique can be seen from the high frequency end ofthe spectrum as shown in Fig. 14 and is approximately3.2 amps. This results in a negative sequence current ofapproximately .25 amps which is the input to the observeras shown in Fig. 4. After the feedback is switched to theback EMF estimates, the high frequency voltage is turnedoff and the only fundamental current remains.

Results from operation at 52,000 rpm are presentedin Figs. 15 through 17. The fundamental frequency isapproximately 1733 Hz. These results were achievedwith a 290 volt dc bus, 40 kHz controller sample rate and65 kHz switching frequency using a MOSFET basedinverter.

0 0.002 0.004 0.006 0.008 0.01time, seconds

Figure 15: PHASE CURRENTFOR 52,000 RPM OPERATION

52 5I

52

51 5

0 0_ % ,oooOI 0_Figure 16: ESTIMATED

ROTORSPEED FOR ACOMMANDEDVALUE OF

52,000 RPM

fr4_00ncy Hz6000

Figure 17: PHASE CURRENTFREQUENCYSPECTRUMFOROPERATIONAT 52,000 RPM

CONCLUSIONSNASA Glenn Research Center has demonstrated

hybrid sensorless control of a surface-mount PM machinefor a flywheel application at speeds up to 52,000 rpm inan in-house test facility. The PM machine is started andaccelerated to 1,500 rpm using the self-sensingtechnique of position estimation in a closed loop speedcontrol with field orientation. Then the feedback is

changed to the back EMF estimates and the additionalhigh frequency voltage necessary for the self-sensingmethod is turned off. The machine is then controlled to

the desired speed using the back EMF estimates. Theself-sensing method is initially calibrated by commandingthe rotor to a known position and determining if an offsetangle of 180 ° is necessary. This hybrid technique allowsthe machine to be operated over its full speed range,from start up to full speed, without a sensor for positionfeedback. This was critical in this application as thismotor was not provided with a resolver or an encoder.

Future work will include automating the transfer fromself-sensing feedback to back EMF feedback at a pre-determined speed. In addition, the exact phase errors forthe different methods will be studied more thoroughlyonce a machine with a resolver is available.

REFERENCES

Jahns, Thomas M. Motion Control with Permanent-Magnet AC Machines, Proceedings of the IEEE, Volume82, No. 8, August, 1994, page 1241-1252.

Jansen, Patrick and Robert D. Lorenz,"Transducerless Position and Velocity Estimation inInduction and Salient AC Machines", IEEE Transactionson Industry Applications, Vol. 31, No. 2, March/April 1995pp. 240-247.

Johnson, James P., M. Ehsani, Y. Guzelgunler,"Review of Sensorless Methods for Brushless DC",Conference Record of the 34th IAS Annual Meeting,1999, Vol.: 1, pp 143 -150.

Kenny, B., P. Kascak, et.al. "Advanced Motor ControlTest Facility For NASA GRC Flywheel Energy StorageSystem Technology Development Unit", NASA TM 2001-210986.

Krishnan, Ramu, Permanent Magnet Synchronous andBrushless DC Motor Drives: Theory, Operation,Performance, Modeling, Simulation, Analysis and Design,Virginia Tech., Blacksburg, Virginia, 1999.

Lorenz, Robert D., K.W. Van Patten, "High-resolutionvelocity estimation for all-digital, AC servo drives" IEEETransactions on Industry Applications, Vol 27, No. 4,July/August 1991, pp. 701-705.

Matsui, Nobuyuki, "Sensorless PM Brushless DCMotor Drives", IEEE Transactions on IndustrialElectronics, Vol. 43, No. 2, April, 1996, pp. 300-308.

Patel, N., T. O'Meara, J. Nagashima, R. Lorenz,"Encoderless IPM Traction Drive for EWHEV's",Conference Record of the 2001 IEEE IndustryApplications Conference, Chicago, II.

Wu, Rusong and Gordon Slemon, "A PermanentMagnet Motor Drive Without a Shaft Sensor", IEEETransactions on Industry Applications, Vol 27, No. 5,September/October 1991, pp. 1005-1011.

NASA/T_2002-211726 7

Form ApprovedREPORT DOCUMENTATION PAGEOMB No. 0704-0188

Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,

gathering and maintaining the data needed, and completing and review#lg the collection of information. Send corrlments regarding this burden estimate or any other aspect of this

collection of information, including suggestions for reducing this burden, to Washington Headquarters Services. Dhectorate for Information Operations and Reports, 1215 Jefferson

Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget Paperwork Reduction Project (0704-.0188), Washington, DC 20503.

1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3, REPORT TYPE AND DATES COVERED

July 2002 Technical Memorandum

5. FUNDING NUMBERS4. TITLE AND SUBTITLE

Sensorless Control of Permanent Magnet Machine for NASA Flywheel


& AUTHOR(S)

Barbara It. Kenny and Peter E. Kascak

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES)

National Aeronautics and Space Administration

John H. Glenn Research Center at Lewis Field

Cleveland, Ohio 44135 - 3191

WU-755-1A-09-00

8. PERFORMING ORGANIZATIONREPORT NUMBER

E----13477

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORINGAGENCY REPORT NUMBER

National Aeronautics and Space Administration

Washington, DC 20546- 0001 NASA TM--2002- 211726IECEC-2002-20072

11. SUPPLEMENTARY NOTES

Prepared for tile 37th Intersociety Energy Conversion Engineering Conference sponsored by the Institute of Electrical and

Electronics Engineers, Electron Devices Society, Washingtom DC, July 28-August 2, 2002.. Barbara H. Kenny, NASA

Glenn Research Center, and Peter E. Kasc_, Ohio Aerospace Institute, Brook Park, Ohio 44142. Responsible person,

Barbara tt. Kenny, organization code 5450, 2.16-433-6289.

12a, DISTRIBUTION/AVAILABILITY STATEMENT

Unclassified - Unlimited

Subject Categories: 20 and 44 Distribution: Nonstandard

Available electronically at Ntp://gltrs.grc.n;_sa.gov/GIiI'RS

"l-his publication is available from the NASA Center for AeroSpace In_brmation, 301-621-0390.

12b. DISTRNBUTION CODE

13. ABSTRACT (Maximum 200 words)

This paper describes the position sensorless algorithms presently used in the motor control for the NASA "in-house"

development work of the flywheel energy storage system. At zero and low speeds a signal il_iection technique, the

self-sensing method, is used to determine rotor position. At higher speeds, an open loop estimate of lhe back EMF of tile

machine is made to determine the rotor position. At start up, tile rotor is set to a known position by commanding dc into

one of tile phase windings. Experimental results up to 52,000 rpm are presented.

14. SUBJECT TERMS

Flywheel energy storage; Sensorless control; Permanent magnet machine; Motor drive

17. SECURITY CLASSIFICATIONOF REPORT

Unclassified

NSN 7540-01-280-5500

15. NUMBER OF PAGES

1316. PRICE CODE

18, SECURITY CLASSiFiCATiON 19. SECURITY CLASSiFiCATiON 20. LiMiTATiON OF ABSTRACTOF THIS PAGE OF ABSTRACT

Unclassified Uncl assifi ed

Standard Form 298 (Rev. 2-89)

Prescribed by ANSI Std. Z39-18298-102

Documents

Sensorless Control of Permanent Magnet Machine for NASA